Improve topic generation by excluding unnecessary phrases

Enabled for Public preview General availability
Admins, makers, or analysts, automatically This feature is released. Jun 11, 2020 -

Business value

Analyzing support topic clusters helps businesses optimize support delivery and improve customer experience, but can be time-consuming. This feature helps organizations improve the quality of AI-driven clustering by identifying phrases, such as a company name, that can skew the clustering and should be excluded.

Feature details

Dynamics 365 Customer Service Insights automatically parses support case data and groups semantically similar cases into autogenerated topics by using natural language understanding. Knowing which are the top-trending topics allows managers to optimize their support and provide feedback to core product teams. However, common phrases like the company or product name can dilute the strength of the topic-clustering algorithm, making the support topic clusters less useful. Today, users can specify delimiters that are used as tags (among others) within the case title that can be ignored to help alleviate this problem, but there's no way to ignore non-delimited text.

With this release, users will be able to specify additional keywords or phrases (such as the company name or other common, irrelevant terms) that should be ignored when the text is parsed, resulting in more accurate and meaningful topic generation.

See also

Improve data quality by cleaning support case data (docs)